feature-engineering-and-feature-selection/A Short Guide for Feature Engineering and Feature Selection.md at master · Yimeng-Zhang/feature-engineering-and-feature-selection · GitHub

Feature engineering and selection is the art/science of converting data to the best way possible, which involve an elegant blend of domain expertise, intuition and mathematics. This guide is a concise reference for beginners with most simple yet widely used techniques for feature engineering and selection. Any comments and commits are most welcome. The field of Machine Learning seeks to answer the question "How can we build computer systems that automatically improve with experience, and what are the fundamental laws that govern all learning processes?" Narrowly speaking, in data mining context, machine learning (ML) is the process of letting computers to learn from historical data, recognize pattern/relationship within data, and then make predictions. There can be many ways to divide the tasks that make up the ML workflow into phases. But generally the basic steps are similar as the graph above. Definition: any measurable property/characteristic of a phenomenon being observed. They are called'variables' because the value they take may vary (and it usually does) in a population. Note: In reality we may have mixed type of variable for a variety of reasons. For example, in credit scoring "Missed payment status" is a common variable that can take values 1, 2, 3 meaning that the customer has missed 1-3 payments in their account. And it can also take the value D, if the customer defaulted on that account. We may have to convert data types after certain steps of data cleaning.